Session S84.5

An Electrophysiological Cardiac Model Approach to Measuring T-Wave Alternans

MA Mneimneh, RJ Povinelli*

Marquette University
Milwaukee, WI, USA

The 2008 Computers in Cardiology Challenge is to automatically identify and measure T-wave alternans, which is an important clinical indicator for sudden cardiac death. This study applies an electrophysiological cardiac model to the problem of characterizing the T-wave variability. Thus, the hypothesis is that the existence and magnitude of T-wave alternans can be identified and measured using a cardiac inverse problem approach, where the magnitude of the alternans are measured in the model space. The dataset used in this study is a collection of records from the following Physionet databases: Long-Term ST, PTB Diagnostic ECG, MIT-BIH ST Change, Sudden Cardiac Death Holter, and BIDMC Congestive Heart Failure. Additionally, a simulated ECG dataset is used to study the sensitivity and specificity of the proposed approach under various noise conditions. To address the hypothesis, an electrophysiological cardiac model, based on characterizing six key cardiac regions, is applied to the T-waves in each record. An inverse solution is generated for each T-wave. The difference between two T-waves is measured in the inverse solution space. The magnitude of the T-wave alternans is given as the average of all consecutive differences in a record. Preliminary results on the simulated ECG data set show that the approach is able to differentiate between 5, 10, 20, and 100 microvolt T-wave alternans in the presence of Gaussian noise of 0, 5, 10, and 20dB. The score from the challenge, which is the Kendall rank correlation coefficient, is 0.41. The score measures the consistency of the rank ordering of our approach’s generated T-wave alternans measurements with a median rank ordering generated from all entries. This is done to allow comparisons of T-wave alternans analysis approaches, which may generate different T-wave scores.

(Abstract Control Number: 390)